Wenan Chen

Automated ventricular systems segmentation in brain CT images by combining low-level segmentation and high-level template matching

BackgroundAccurate analysis of CT brain scans is vital for diagnosis and treatment of Traumatic Brain Injuries (TBI). Automatic processing of these CT brain scans could speed up the decision making process, lower the cost of healthcare, and reduce the chance of human error. In this paper, we focus on automatic processing of CT brain images to segment and identify the ventricular systems. The segmentation of ventricles provides quantitative measures on the changes of ventricles in the brain that form vital diagnosis information.MethodsFirst all CT slices are aligned by detecting the ideal midlines in all images. The initial estimation of the ideal midline of the brain is found based on skull symmetry and then the initial estimate is further refined using detected anatomical features. Then a two-step method is used for ventricle segmentation. First a low-level segmentation on each pixel is applied on the CT images. For this step, both Iterated Conditional Mode (ICM) and Maximum A Posteriori Spatial Probability (MASP) are evaluated and compared. The second step applies template matching algorithm to identify objects in the initial low-level segmentation as ventricles. Experiments for ventricle segmentation are conducted using a relatively large CT dataset containing mild and severe TBI cases.ResultsExperiments show that the acceptable rate of the ideal midline detection is over 95%. Two measurements are defined to evaluate ventricle recognition results. The first measure is a sensitivity-like measure and the second is a false positive-like measure. For the first measurement, the rate is 100% indicating that all ventricles are identified in all slices. The false positives-like measurement is 8.59%. We also point out the similarities and differences between ICM and MASP algorithms through both mathematically relationships and segmentation results on CT images.ConclusionThe experiments show the reliability of the proposed algorithms. The novelty of the proposed method lies in its incorporation of anatomical features for ideal midline detection and the two-step ventricle segmentation method. Our method offers the following improvements over existing approaches: accurate detection of the ideal midline and accurate recognition of ventricles using both anatomical features and spatial templates derived from Magnetic Resonance Images.

Segmentation of ventricles in brain CT images using Gaussian Mixture Model method

In this paper, a segmentation method using Gaussian Mixture Model (GMM) combined with template match is proposed for analysis of brain CT images. The specific aim of this method is to extract ventricles from brain CT images. These can then be used for automated detection of the midline shift in brain. In the method, different types of brain tissue, of which the ventricles form the region of interest, are segmented using multiple Gaussian mixtures. Expectation Maximization (EM) method is used to train the GMM. Ventriclular tissue is then detected in the segmented regions using template matching. Other segmentation methods, including K-means clustering and Iterated Conditional Modes (ICM), are also implemented and their results are compared with those of the proposed method. The algorithms are evaluated against a dataset of brain CT images captured from both normal and TBI cases. The segmentation results show the advantages of the proposed GMM-based method for brain tissue modeling. The computational complexity of the proposed method is also discussed, as well as the means to address this issue. The proposed GMM-based method allows accurate segmentation of ventricles required for detection of the shift in the midline.

Automated segmentation of lateral ventricles in brain CT images

It is estimated that every year, 1.5 million people in the United States sustain a traumatic brain injury (TBI). Over 50,000 of these patients will not survive, and many others will be left permanently disabled. TBI is known to be accompanied by an increase in intracranial pressure (ICP), as the presence of hematomas compresses the brain tissue. Severe ICP can be fatal, and so must be monitored. This typically requires cranial trepanation, a risky procedure for the patient. However, some signs of increased ICP are visible on medical scans. For example, the lateral ventricles may change in size and position, depending on the location of the original injury. In this paper, we focus on automatic processing of CT brain images to segment and identify the lateral ventricles, using both iterated conditional models (ICM) and maximum a posteriori spatial probability (MASP). The ideal midline of the brain is found via exhaustive search based on skull symmetry and tissue features. The horizontal shift in the ventricles associated with increased ICP can then later be calculated based on the ideal midline. The novelty of the proposed method lies in its combination of anatomical features with template matching against MRI images, its stepwise improvement of the detected actual midline, and its comparison of two existing methods, ICM and MASP, for ventricle detection. The relatively large size of the CT dataset used for testing increases the reliability of the results.

Agent based modeling of blood coagulation system: Implementation using a GPU based high speed framework

The coagulation and fibrinolytic systems are complex, inter-connected biological systems with major physiological roles. The complex, nonlinear multi-point relationships between the molecular and cellular constituents of two systems render a comprehensive and simultaneous study of the system at the microscopic and macroscopic level a significant challenge. We have created an Agent Based Modeling and Simulation (ABMS) approach for simulating these complex interactions. As the scale of agents increase, the time complexity and cost of the resulting simulations presents a significant challenge. As such, in this paper, we also present a high-speed framework for the coagulation simulation utilizing the computing power of graphics processing units (GPU). For comparison, we also implemented the simulations in NetLogo, Repast, and a direct C version. As our experiments demonstrate, the computational speed of the GPU implementation of the million-level scale of agents is over 10 times faster versus the C version, over 100 times faster versus the Repast version and over 300 times faster versus the NetLogo simulation.

Intracranial pressure level prediction in traumatic brain injury by extracting features from multiple sources and using machine learning methods

This paper proposes a non-intrusive method to predict/estimate the intracranial pressure (ICP) level based on features extracted from multiple sources. Specifically, these features include midline shift measurement and texture features extracted from CT slices, as well as patients demographic information, such as age. Injury Severity Score is also considered. After aggregating features from slices, a feature selection scheme is applied to select the most informative features. Support vector machine (SVM) is used to train the data and build the prediction model. The validation is performed with 10 fold cross validation. To avoid overfitting, all the feature selection and parameter selection are done using training data during the 10 fold cross validation for evaluation. This results an nested cross validation scheme implemented using Rapidminer. The final classification result shows the effectiveness of the proposed method in ICP prediction.

Automated midline shift and intracranial pressure estimation based on brain CT images

In this paper we present an automated system based mainly on the computed tomography (CT) images consisting of two main components: the midline shift estimation and intracranial pressure (ICP) pre-screening system. To estimate the midline shift, first an estimation of the ideal midline is performed based on the symmetry of the skull and anatomical features in the brain CT scan. Then, segmentation of the ventricles from the CT scan is performed and used as a guide for the identification of the actual midline through shape matching. These processes mimic the measuring process by physicians and have shown promising results in the evaluation. In the second component, more features are extracted related to ICP, such as the texture information, blood amount from CT scans and other recorded features, such as age, injury severity score to estimate the ICP are also incorporated. Machine learning techniques including feature selection and classification, such as Support Vector Machines (SVMs), are employed to build the prediction model using RapidMiner. The evaluation of the prediction shows potential usefulness of the model. The estimated ideal midline shift and predicted ICP levels may be used as a fast pre-screening step for physicians to make decisions, so as to recommend for or against invasive ICP monitoring.

Wavelet based analysis of physiological signals for prediction of severity of hemorrhagic shock

Rapid detection and treatment of hemorrhagic injuries are important factors in decreasing mortality in the battlefield and civilian trauma settings. In this study, novel features based on discrete wavelet transformation (DWT) were used to analyze physiological signals for prediction of central hypovolemia severity in humans. These features were defined based on approximate and detailed DWT coefficients extracted from physiological signals such as the electrocardiogram (ECG), arterial blood pressure (ABP), and thoracic impedance (IZT and DZT) signals, collected on healthy humans exposed to a hemorrhage model called lower body negative pressure (LBNP). The LBNP protocol consisted of applying 0, -15, -30, -45, -60, -70 mm Hg pressure to the lower half of the body, for 5 minutes at each stage. These LBNP levels were divided into three classes: mild, moderate, and severe. Machine learning algorithms were applied to predict the severity of blood loss based on the features extracted from the physiological signals. One of the objectives of this study was to compare the utility of using multiple physiological signals in prediction of the severity of hypovolemia as opposed to only using ECG. The classification results indicate that SVM has the highest accuracy at 82%. SVMs average precision and recall for all three classes are 79.2% and 79.8%, respectively. This shows that the wavelet-based method using multiple signals has the ability of rapidly determining the degree of volume loss, providing a potential tool for real-time remote triage and decision making in victims of trauma.

Predictability of intracranial pressure level in traumatic brain injury: features extraction, statistical analysis and machine learning-based evaluation

This paper attempts to predict Intracranial Pressure (ICP) based on features extracted from non-invasively collected patient data. These features include midline shift measurement and textural features extracted from Computed axial Tomography (CT) images. A statistical analysis is performed to examine the relationship between ICP and midline shift. Machine learning is also applied to estimate ICP levels with a two-stage feature selection scheme. To avoid overfitting, all feature selections and parameter selections are performed using a nested 10-fold cross validation within the training data. The classification results demonstrate the effectiveness of the proposed method in ICP prediction.

Texture analysis of brain CT scans for ICP prediction

Elevated Intracranial Pressure (ICP) is a significant cause of mortality and long-term functional damage in traumatic brain injury (TBI). Current ICP monitoring methods are highly invasive, presenting additional risks to the patient. This paper describes a computerized non-invasive screening method based on texture analysis of computed tomography (CT) scans of the brain, which may assist physicians in deciding whether to begin invasive monitoring. Quantitative texture features extracted using statistical, histogram and wavelet transform methods are used to characterize brain tissue windows in individual slices, and aggregated across the scan. Support Vector Machine (SVM) is then used to predict high or normal levels of ICP using the most significant features from the aggregated set. Results are promising, providing over 80% predictive accuracy and good separation of the two ICP classes, confirming the suitability of the approach and the predictive power of texture features in screening patients for high ICP.

Actual Midline Estimation from Brain CT Scan Using Multiple Regions Shape Matching

Computer assisted medical image processing can extract vital information that may be elusive to human eyes. In this paper, an algorithm is proposed to automatically estimate the position of the actual midline from the brain CT scans using multiple regions shape matching. The method matches feature points identified from a set of ventricle templates, extracted from MRI, with the corresponding feature points in the segmented ventricles from CT images. Then based on the matched feature points, the position of the actual midline is estimated. The proposed multiple regions shape matching algorithm addresses the deformation problem arising from the intrinsic multiple regions nature of the brain ventricles. Experiments on the CT scans from patients with traumatic brain injuries (TBI) show promising results, particularly the proposed algorithm proves to be quite robust.